| Policy_Number | Company_Sector | Company_Name | Premium | Retention |
|---|---|---|---|---|
| GL00001 | Technology | Technology Co 2 | $40,404 | Medium |
| GL00002 | Technology | Technology Co 3 | $41,952 | Low |
| GL00005 | Healthcare | Healthcare Co 6 | $32,854 | Medium |
| GL00007 | Construction | Construction Co 8 | $26,090 | High |
| GL00008 | Healthcare | Healthcare Co 9 | $25,467 | Medium |
| GL00009 | Renewable Energy | Renewable Energy Co 10 | $81,331 | Medium |
| GL00012 | Technology | Technology Co 13 | $37,076 | High |
| GL00013 | Renewable Energy | Renewable Energy Co 14 | $68,826 | Low |
| GL00014 | Renewable Energy | Renewable Energy Co 15 | $50,257 | High |
| GL00015 | Renewable Energy | Renewable Energy Co 16 | $34,535 | High |
Randomised Portfolios in Insurance
Mick Cooney mcooney@describedata.com
Introduction
Basic Concept
Insurance Use Cases
Random Portfolios
Risk/Reward Tradeoff
Risk Tolerance
Risk Limits
Only NASDAQ, NYSE listed stocks
No single stock greater than 5% of portfolio
Market Cap greater than 100M USD
No sector greater than 25% of portfolio
Key Finding
Portfolio performance is highly dependent on risk limits
Separate “alpha” from “beta”?
CAPM Model
\[ r = \alpha + \beta \, m \]
Basic Approach
- Generate random portfolios obeying risk limits
- Measure performance of random portfolios
- Compare actual with simulated returns
But what about insurance???
Steal the simple idea…
Use random portfolios of policies to do stuff
What sort of stuff?
Insurance Use Cases
Premium Forecasting
Lloyd’s Underwriter
Premium Planning
Knows the Market
Assigned High, Medium, Low chances of the retaining policy
Assign probability for policy
Run simulations
Premium Planning
US Liability Business
Grow from 60mm USD to 80mm USD
Where does the premium come from?
Northeast US -> 10mm
Heathcare sector -> 5mm
Small business -> 5mm
Co-occurrence of Risk Factors
Allocate by simulating policies
Also allocates across other policies
Portfolio Optimisation
Involved
Basic visualisations
Use pricing / loss models
Simulate portfolios
What does ‘optimal’ mean?
Proper use may require rethinking relationships
Summary
Random Portfolios are a powerful idea
Lots of use cases in insurance
Requires computational power
Thank You!